Claude Security: Control Claude and Claude Code Use
Log prompts from claude.ai, the Claude desktop app and the Claude Code CLI. Separate personal accounts and protect code and secrets.
- Claude
- ChatGPT
- Microsoft Copilot
- Google Gemini
- Allow & log
- Redact
- Block
- Deny app
Risks of Claude use
Consumer plans
since 2025, Anthropic uses chats on Free, Pro and Max plans for model training if the user allows it. The user controls this setting.
Agent tools
Claude Code runs in the terminal, reads files and runs commands. Secret files in the repository can enter the context.
Large uploads
Claude accepts long documents. One upload can carry a full contract or data export.
Recommended AI Control policy
Scope: Windows, macOS and Linux devices with the Moftgate agent. Personal phones without the agent are out of scope.
- 01
Browsers and Claude desktop × Claude: Allow & log.
- 02
CLI application type × Claude API: Block on the secret key classification.
- 03
When customer or personal data classifications match: Redact.
Allow & log
The prompt is sent and recorded.
Redact
Matched data is masked and the masked prompt is sent.
Block
The prompt is not sent and the event is recorded.
Deny app
The app cannot talk to this provider at all.
Questions about Claude
Details are based on policies published by Anthropic and public reports. Plans and settings can change; check current terms with the vendor.
See your Claude use in one week
AI Control logs prompts and masks or blocks sensitive data. Data stays on your own servers.